Using a simple least squares method to fit the scores of all NFL games  (discussed in Chapter 40 of my book Mathletics) here are the top 5 teams:

1. Saints

2. Colts

3. Broncos

4. Patriots

5. Jets.

Best Offenses Saints and Patriots

Best Defenses  Broncos and Jets

Projected FInal Season Records

• Arizona 12-4
• Falcons 11-5
• Ravens 9-7
• Bils  7-9
• Panthers 3-13
• Bears 8-8
• Bengals 11-5
• Browns 4-12
• Cowboys 10-6
• Broncos 14-2
• Lions3-13
• Packers 10-6
• Colts 14-2
• Jaguars6-10
• Dolphins 9-7
• Vikings 12-4
• Patriots 11-5
• Saints 15-1!!
• Giants 10-6
• Jets 10-6
• Raiders 4-12
• Eagles 8-8
• Steelers 10-6
• Rams 1-15
• Chargers 9-7
• 49ers 9-7
• Seahawks 8-8
• Tampa Bay 1-15
• Titans 3-13
• Redskins 3-13

Using a logistic regression model and adjusting for the fact that Yankees and Phillies will not use some of their regular season pitchers the Yankees have a 65%  chance to win series and Phillies a 35% chance. Here are more details:

• 9.6% of time Yankees in 4
• 14.5% of time Yankees in 5
• 21.7% of time Yankees in 6
• 19.2% of time Yankees in 7
• 3.4% of time Phillies in 4
• 9.8% of time Phillies in 5
• 10.4 of time Phillies in 6
• 11.3% of time Phillies in 7

The simulations were done with Palisade’s Monte Carlo Excel -add-in @RISK.

A key to football decision-making  (as detailed in my book Mathletics) is to assign a point value to each down, yard line and yards to go for first down situation. There are nearly 10,000 of these situations. In the Wall Street Journal (September 2009) we used this idea to show how valuable Michael Vick’s rushing offense was during 2006. Here is a brief history of point values in football. Let me know if I have left anything out of importance in this chronology.

1. Carter and Machol 1971 derived points values for first and 10 situations in their Operations Research article.
2. Cabot, Sagarin and Winston in 1983 derived point values for each down, yards to go and yardline situation. This was submitted to Operations Research and was discussed in a September 15, 1983 Bloomington Herald -Telephone  article.  See http://www.kiva.net/~jsagarin/sports/wham_bam.pdf.

3. Pete Palmer (1989) in his excellent book Hidden Game of Football  discussed the concept of 1st down and 10 values and indicated how they might be used for football decision-making

4.   In 2003 the website Footballoutsiders.com derived their own 1st down and 10 point values and estimated values for other situations and these values were used to rate players, team offenses and team defenses.

5. In 2006 economist David Romer published in the Journal of Political Economy an article that used his point values for 1st down and 10 situations to show that teams attempt punts and field goals much more than should.

6. In September 2009 Levitt  (of Freakonomics fame) and Kovash derived values for every down, yards to go and yard line situation and used these values to analyze football decision making.

• Here is Carmelo Anthony’s Adjusted +/-  listed in chronological order: -6, 3 1, -2, -2, 7. So Carmelo finally got it last year, Most of this improvement was at the defensive end.
• Here is Troy Murphy’s: -2, -3, -8, -7, -8, -4, -10, -1, +1. So in Troy’s two years with the Pacers he iproved a great deal.
• Here is the Adjusted +/- sequence for the most underrated player of the decade: Brad Miller (as an IU professor it kills me to say this) +9 , +8, +4, +6, +8, +4. Amazing consistency and quality play!
• Another underrated player: Darius Songaila. Here is his sequence +8, +8, +2,+1, +5,+2. Of course, he plays limited minutes, but he always seems to help his team.
• Another player who eventually  ”got it”: Danny Granger. -6, - 3 ,-3 +6!
• Another underrated player Pachulia: +2, 0, +1, 0 and -1.

As Joe Friday said on Dragnet, “Just the facts maam.”  Here are some facts about the Thunder.

• Based on KD’s 2 years in the league most adjusted +/- numbers have him rated around -7 points. Even with the noisiness in these estimates there is less than a 5% chance that the true value of his adjusted +/- for those two years is above average.

Let’s break down all Thunder minutes last year into 3 lineup combinations and look at how (adjusting for strength of opposition) the Thunder played. The standard deviation of these estimates (rounded off) is also given

• Collison Westbrook Green and KD in +.4 points (std dev 4 points)
• All other KD minutes -11.2 (st dev 2 points)
• All minutes with KD out -2.6 (std dev 3 points)

Draw your own conclusions from this information.

Who contributed more to the Thunder’s (limited) success in 2008-2009: Kevin Durant or Russell Westbrook? Most NBA fans would say clearly Kevin Durant. As Henry Abbott courageously pointed out on True Hoop http://myespn.go.com/blogs/truehoop/0-45-42/Memo-to-a-Young-Baller.html, this may not be the case. Looking over all the lineup combinations used by the Thunder during 2008-2009 our adjusted +/- system found Westbrook to be better than Durant. How can this be? Some intuition can be found by looking at the following tables. The numbers show per 48 minutes (after adjusting for the ability of opponents faced, how well the Thunder played in several situations. From the first table we find that with Westbrook and Green in, for example, the Thunder played 4.03 points worse than average while with Durant and Collison in the Thunder played 8.63 points worse than average.

The next table tells us, for example, that with Durant out and Krstic in the Thunder played 5.68 points better than average, but with Krstic in and Westbrook out the Thunder played 4.06 points worse than average.

Row Player in Column Player Out

Note that in most cases, the Thunder played better with Durant out relative to Durant in and better with Westbrook in compared to Westbrook out.

This table shows how the Thunder did with both players out. For example, when Westbrook and Krstic were out the Thunder played 12.92 points worse than average.

Both Out

I know we should look at holding the other players constant, like Green Westbrook and Collison with or without Durant etc. That is what Adjusted +/- does!!!!!!! If you can look at these numbers and see how Durant contributed more to the Thunder’s success last year than Westbrook, please let me know what I am missing. Thanks.

Of course, these numbers are based on lineup combinations used by the Thunder, but what else can we do?

After updating my logistic regression model to include the divisional playoffs here are my updated predictions. Note that  we correctly predicted all 4 divisional series.

1. Yankees 55% chance to beat Angels
2. Dodgers 54% chance to beat Phillies

Chances to Win World Series

1. Yankees 38%
2. Angels 31%
3. Dodgers 18%
4. Phillies 13%

Mark Cuban decided not to rehire our statistical analysis team (WINVAL) this season. We greatly enjoyed our time with the Mavs.  Mark is a great person and a terrific owner. The Mavs never won that elusive championship, but during our tenure with the Mavs only the Mavs and Spurs won at least 50 games each year.  We’d like to think that we played some small role in that success.

    As the season progresses, we will blog on what makes  an NBA teams win or lose, and how each team can improve their performance. It would have been fun last year, for example, to point out before the Bulls Celtics series that Brad Miller with 3 guards was by far the Bulls best lineup. This proved to be true in the great Bulls-Celtics series. It would also have been fun to point out that after two or three games the lineups that were hurting the Cavs against the Magic. These lineup combinations continued to hurt the Cavs throughout the series.

   We wish the Mavs well, and our goal throughout the season will be to show how data driven analysis to can improve the performance of a business (in this case an NBA team). As the great statistician Edwards Demings said “In God we trust, all others need data.” Of course data is only one  of the inputs needed for a sound decision making process, but I am sure we can show  how  sound analysis of data leads to valuable insights that can drive improved performance in any organization.

A Mathletic Look at Statistics in Sports, including Forecasts for the Redskins and Ravens

Contact: Mark Schoeff Jr., chair, NPC Newsmakers Committee (202-662-7218,  [email protected] );  Lura Forcum, Kelley School of Business at Indiana University (812-856-1232, [email protected])

Indiana University Professor Wayne Winston will discuss sports and statistics at a National Press Club Newsmakers press conference at 10 a.m. on Tuesday, October 20, 2009 in the Lisagor Room, 13th floor, National Press Building, 529 14th St., N.W.  Winston will explain how statistics can be used to predict how a player changes a team’s chances of winning. He’ll also reveal his stats-based projections for the upcoming Redskins and Ravens seasons.

 Winston is the author of Mathletics (Princeton University Press), a new book that explains why the sports statistics that newspapers publish are useless and which stats actually matter when evaluating players and teams in football, basketball and baseball. Winston is a professor at the Kelley School of Business at Indiana University.
Winston will be joined by Timothy A. Franklin, director of the National Sports Journalism Center at Indiana University and former editor of The Baltimore Sun and sports editor of the Chicago Tribune. Franklin will discuss how sports statistics can enable reporters to give their readers new insights into the teams they cover. He also will address trends in sports media more generally.

Formed in 2009, the National Sports Journalism Center is based at Indiana University’s School of Journalism. The center’s mission is to provide instruction, discussion and research on the issues facing sports journalists. The center is now the home of the Associated Press Sports Editors, the nation’s largest professional sports media organization. It also produces the nation’s most comprehensive sports media Web site, www.sportsjournalism.org. 

Russ Thaler, anchor for SportsNite on Comcast SportsNet as well as chief digital correspondent at CSNWashington.com, will moderate the event. Thaler was formerly the host of Washington Post Live. He now blogs about the Redskins, Ravens, and sports in general when he’s not anchoring SportsNite.

It saddens me that Andre Iguodala did not make the All-Star Team last year. Using Adjusted +/- we have him as the NBA’s 7th best player in 08-09. We estimate he  is 11.46 points better than an average player. He is 3.64 points better than average and 7.82 points better than average on defense. I believe most followers of the NBA do not fully appreciate AI’s defensive contributions. Below we see how well the 76ers played last year with AI in and AI out paired with other players. For example, with Young (another good player) in and AI in Philly was 5.36 points better than average but with Young in and AI out Philly played 8.59 points worse than average. Adjusted +/- is simply a weighted average of what AI contributed after adjusting for the abilities of the other 9 guys on the court. How could AI not make the all-star team? I do not understand!